Host patching with Citrix and 3D GPU


First, I’ll say that i’m primary a Citrix admin, with some vMware experience.

We have a vMware 6.5 environment where I work. We provide non persistent pooled desktops with nVidia Grid GPU cards backed by nutanix. Due to the presence of these hardware GPU cards, machines can’t be v Motioned. We have 15 hosts and around 150 VMs.

We have enough host capacity to take a few out for patching, but can’t figure out how to get the machines off of them. This means we all get to work on the weekend and causes and outage.

From the Citrix side, I can’t control what hosts a machine starts on. Every time the machines restart (after users logoff), it may start on the same host, or may start on a different host.

My hope was to setup some sort of rolling update.

We would put some hosts in a maintenance mode of sorts. So that on restart, machines would automatically start on a different host. I’m just not sure the best way to accomplish this. I’ve thought about using affinity rules and am comfortable in powershell.

Anyone have any solid approaches to this?

Starting State:

HOST 1: VM 1,2,3

HOST 2: VM 4,5,6

HOST 3: VM 7,8,9

Action: Put HOST 1 out of circulation

time… VM 1 and VM3 restart due to user logoff

HOST 1: <Now Patching>

HOST 2: VM 4,5,6, 1, 2

HOST 3: VM 7,8,9, 3

Action: Put host back into circulation, pull HOST 2 out

View Reddit by anteck7View Source

Related Articles


  1. Few questions & options.

    1) Are you a 24×7 shop? If not, do a weekly shutdown of all desktop VMs on like Saturday night, combined with a VUM update on the cluster. Solved.

    2) Are you able to update VMware? vSphere 6.7u1 supports vMotion with GPUs.

    3) You could probably do it using tags and categories with HA. Create a tag called “Active GPU Host” and limit HA to only starting on a host with that tag. Then as part of your process you remove the tag from a host, that night when all VMs are off due to logoff circulation, no VMs will start on it.

  2. I generally update outside of business hours. Power them all down and patch ESXi. I’m looking to upgrade to 6.7u3 to avoid that headache.

Leave a Reply